disasters, lessons learned, and fantasy documents
TRANSCRIPT
Disasters, Lessons Learned, andFantasy Documents
Thomas A. Birkland
School of Public and International Affairs, North Carolina State University, Campus Box 8102, Raleigh,NC 27511, USA. E-mail: [email protected]
This article develops a general theory of why post-disaster ‘lessons learned’ documents
are often ‘fantasy documents’. The article describes the political and organizational
barriers to effective learning from disasters, and builds on general theory building on
learning from extreme events to explain this phenomenon. Fantasy documents are not
generally about the ‘real’ causes and solutions to disasters; rather, they are generated to
prove that some authoritative actor has ‘done something’ about a disaster. Because it is
difficult to test whether learning happened after an extreme event, these post-disaster
documents are generally ignored after they are published.
1. Introduction
A staple of crisis management and emergency
response is the post-response report, often
known as an ‘after action’ report or a ‘lessons learned’
document. Many of these reports are the routine
product of organizational self-evaluation and are pri-
marily concerned with operational or ‘tactical’ matters.
Indeed, this sort of learning is known to organizational
theorists as ‘single-loop learning’ (Argyris & Schon,
1996), which has very important implications for crisis
management (see Moynihan’s and Deverell’s papers in
this symposium). But I am more concerned with the
second loop, as it were, of ‘double-loop learning’, which
involves learning about the fundamental assumptions
behind policy design at the strategic level. Here, the
claims of ‘lessons’ and ‘learning’ have significant implica-
tions for the supposed lesson learners and the broader
policy system.
Because social and political pressures to create such
lessons learned reports are the greatest in the immedi-
ate aftermath of the event, while the event’s status on
the agenda is freshest, a great deal of attention is paid to
ensuring that lessons really are learned, so that the
worst effects of the next disaster can be avoided.
These pressures also mean that lessons learned
reports are usually very quickly generated. It is difficult
to claim that any actual learning occurred because
insufficient time has elapsed between the event, the
creation of the report, and any subsequent tests of the
‘lessons’. Instead, these documents really focus on
‘lessons observed’ or, more simply, the observations
that officials and experts made about the preparations
before and responses to the crisis or disaster. Moreover,
most of the time, these reports are narrow-bore efforts
to derive meaning for a particular constituency; in the
disaster field, these groups include first responders,
communications experts, and public health officials.
There are few comprehensive efforts to learn broader
strategic lessons about the events based in sound
science; this is consistent with the idea that single-loop
learning is more common than double-loop learning.
In this article, I borrow concepts and terminology
from Lee Clarke, who coined the term ‘fantasy docu-
ments’ (Clarke, 1999). I call many lessons learned
documents ‘fantasy learning documents’ for the same
reason that Clarke terms many pre-disaster plans ‘fan-
tasy documents’: because they are created and dissemi-
nated for rhetorical purposes, even if their authors
somehow believe that learning has really occurred.
To begin, I review the theories of focusing events and
outline a theory of learning from focusing events. I then
develop a general theory of why post-disaster lessons
learned documents are fantasy documents. This is not
true in all cases, of course, but the general trend is
towards producing such documents to prove that some
& 2009 Blackwell Publishing Ltd.
Journal of Contingencies and Crisis Management Volume 17 Number 3 September 2009
authoritative actor has ‘learned its lessons’ about a
disaster and that, given this learning, will not replicate
its errors.
2. Overview and definitions
John Kingdon (1995) uses the term ‘focusing event’ in
his study of agenda setting and alternative selection to
describe a class of political phenomena that can cause
an issue to gain attention in the media and among
various institutions. In my work (Birkland, 1997, 1998,
2006), I further defined focusing events as events that
are sudden, that are known to policy makers and elites
simultaneously, that affect a community or a community
of interest, and that do actual harm, or that suggest the
possibility of greater future harm. My definition of the
term ‘focusing event’ is influenced by Cobb and Elder’s
(1983) work on agenda setting, in which they call
phenomena like focusing events ‘circumstantial reac-
tors’, and Baumgartner and Jones’s (1993) work on the
‘punctuated equilibrium’ model of the policy process,
in which public policies remain rather stable until
something upsets the system’s equilibrium, yielding
change. All of these works acknowledge that sudden
events are important examples of agenda drivers, but
do not go further than that. My work sought to sharpen
the idea of focusing events1 and in showing how
focusing events do not influence all policy domains in
the same way. On the other hand, my definition of
focusing events is rather more restrictive than King-
don’s; this definitional difference will not be resolved
here, but it is important to acknowledge.
Focusing events, by elevating issues on the agenda,
can, says Kingdon, open a ‘window of opportunity’ for
policy change. This window of opportunity can yield
immediate policy change, improved understanding
of the social or the natural forces that lead to a disaster,
or can be an opportunity for a variety of actors to learn
how better to argue for their policy or political
interests. Of course, these outcomes are not mutually
exclusive, and this knowledge can be accumulated
and applied after later focusing events or other change
opportunities. Peter May (1992) defines these three
types of learning as instrumental policy learning, social
policy learning, and political learning. Instrumental
policy learning involves learning about the effectiveness
of various policy tools applied to problems. Social policy
learning relates to learning about the social construc-
tion of problems and the interaction of policies with the
targets of policies. Political learning involves learning
about the effectiveness of rhetorical appeals for policy
change, and involves political strategies and tactics at
the ideological level, rather than the specifics of public
policies. This paper will be mostly concerned with
social and policy learning, although politics and political
learning are undeniably important.
Natural disasters, industrial accidents, and acts of
terrorism – what are together called ‘extreme events’ –
constitute one type of focusing event that can have local
and distant social and political effects. hurricane Katrina
was a local event for the Gulf Coast, while the distant
impacts of a focusing event are illustrated by the sig-
nificant loss of life in Thailand in the 2004 tsunami. This
disaster killed and injured a great many Swedes on
holiday, the governmental response to which had signifi-
cant consequences for Swedish politics (Naik et al., 2005;
Stromback & Nord, 2006; Widfeldt, 2007).
Because these events are undesirable, humans and
their institutions are presumably interested in mitigat-
ing them or preventing their damages from happening in
the first place. For example, the Air Accidents Investi-
gation Branch in the United Kingdom, and its counter-
part in the United States, the National Transportation
Safety Board, exists to collect a vast amount of infor-
mation on aviation incidents, ranging from minor mis-
haps to catastrophic accidents. The catastrophes are
the more focal events, but from nearly every major
aviation accident we have learned about the causes and
‘cures’ for aviation accidents (Perrow, 1999), such that
aviation safety has made remarkable gains (Cobb &
Primo, 2003).
Because learning from disasters is usually the result of
some sort of intensive investigational and study activity,
learning should not be seen as an outcome or a goal of the
process, but should be considered an ongoing activity
within the policy process. George Busenberg defines the
learning process as ‘a process in which individuals apply
new information and ideas to policy decisions’ (2001).
I accept this definition and suggest that focusing events
can provide that new information, although in a relatively
raw form. For example, the risk of a catastrophic terrorist
attack on the United States was no greater on 12
September 2001 than it was on 10 September, but the
September 11 attacks caused the public and elites to be
much more attentive to the terrorism problem. The
focusing event brings information to the attention of a
broader range of people than normally consider the issues.
However, my definition extends somewhat on Bu-
senberg’s by focusing more on the outcome of learning
than on the process – that is to say, I seek evidence of
some sort of change as a result of the new information,
while Busenberg’s definition only requires the applica-
tion of new information, regardless of the policy
decision. Policy learning can be identified if there is
prima facie evidence of policy changes that are reason-
ably linked to the causal factors that connected the
event under consideration to its harms, and if addres-
sing these factors would be likely to mitigate the
problem (Birkland, 2006). For example, we can say
that policy learning has occurred in the United States
after September 11 through a regulatory requirement
that cockpit doors be kept closed and securely locked
Disasters, Lessons Learned, and Fantasy Documents 147
& 2009 Blackwell Publishing Ltd.
Journal of Contingencies and Crisis Management
Volume 17 Number 3 September 2009
during flight (Airline Industry Information, 2001; World
Airline News, 2001). The new requirement is therefore
clearly a response to the insecurity of cockpits pre-
September 11. However, it is also true that cockpit
intrusions were nothing new, and we can speak of the
failure to learn from less catastrophic, but still worri-
some, episodes of deranged passengers seeking to
enter the flight deck (Air Safety Week, 2000; Richfield,
2000). This is an example of double-loop learning
because a small but fundamental policy change occurred
that transcended the usual regulatory adjustments that
characterized single-loop learning.
However, what looks like policy learning – that is, a
change after some sort of external shock – may not be
learning at all, for at least two reasons; first, the
‘lessons’ that may be ‘learned’ after an event may not
be related to the event at all, but, rather, the lessons
had already been ‘observed’ several times before the
event. That existing knowledge was either not taken up
by those who could have acted, or the knowledge was
available, but policy makers and implementers simply
chose not to act on that new knowledge. Examples of
this include the significant evidence of security pro-
blems in civil aviation well before September 11; it took
September 11 to drive these ideas forward on the
agenda. This is entirely consistent with Kingdon’s idea
that focusing events open the window of opportunity
for the joining of problems with pre-existing solutions,
such as better checkpoint screening, cockpit security,
and the like (Cobb & Primo, 2003; Birkland, 2004, 2006,
Chapter 3). Indeed, the oft-stated lament that ‘it takes a
disaster to change anything’ is entirely consistent with
agenda setting and focusing event theory in a wide
range of fields, from the ongoing financial crisis to
industrial accidents and natural disasters. Moreover, at
least intuitively, we know that ‘big’ events are more
likely to yield policy change than are ‘small’ events.
Second, some policy learning is ‘superstitious’ learning,
which either attempts to use ‘lesson drawing’ from other
places or times, regardless of whether the comparison is
apt (Neustadt & May, 1986), or when, in the urge to ‘do
something’, policies are adopted that have little or
nothing to do with the problem at hand. For example,
after the Columbine school shootings near Denver,
Colorado, in 1999, some policy makers sought to
more closely regulate video games and popular music,
which were said, absent sound scientific information, to
cause the sorts of behaviours that led to this disaster
(Haider-Markel & Joslyn, 2001; Lawrence & Birkland,
2004; Larkin, 2007). While no real social policy or
political learning occurred in this incident, there was
considerable evidence of political learning, in which all
manner of arguments – about popular culture, the
availability of guns, the lack of mental health services,
and so on – were honed and deployed in a battle of ideas
that, ultimately, generated more heat than light.
3. Why are disasters change andlearning opportunities?
Disasters are change and learning opportunities be-
cause they provide an opportunity for close analysis of
the things that happened before the disaster, during the
acute phase of the disaster, and in the recovery period.
The opportunities for learning and change come be-
cause these are extreme events, and therefore gain the
attention that routine events do not. These events gain
a great deal of media attention and, therefore, public
attention. If nothing else, decision makers assume that
what is on the media agenda is also high on the public’s
agenda as well. With public attention comes pressure
to do something about the event. What that ‘some-
thing’ might be is often very murky, because focusing
events not only raise an issue on the agenda; they also
elevate the manifold constructions of the issue on the
agenda. Only those constructions that somehow reso-
nate with the public or elites are elevated, even if these
constructions are, in the causal sense, wrong (Hilgartner
& Bosk, 1988; Lawrence & Birkland, 2004). Thus, after
September 11, there were many ‘new’ problems to be
addressed: border and immigration control, flight train-
ing, airline security, illicit money transfers, emergency
pre-paredness, seaport security, law enforcement, and so
on. Many of these issues were opportunistically advanced
on the agenda by interests who had sought policy change
for years; in other words, the event did not provide new
information, but provided new ways of framing an
existing set of policies to achieve a set of goals (in
particular, the advancement of the political right’s law
enforcement agenda). September 11 was an opportunity
to tie their issue to the new world of ‘homeland security’.
But it is at the ‘do something’ juncture that the
opportunity to learn is manifest, but, given the haste of
the decisions made in the wake of these events, the risk
of superstitious learning – that is, learning without some
sort of attempt to analyse the underlying problem – is
greatest. In some cases, pressure to act is so strong that
action is taken immediately, as was the case of the
enactment of the USA Patriot Act in 2001. This event
broke the pattern in the United States in which most
legislation and regulatory change followed some sort of
investigative or ‘after action’ report (Rubin et al., 2003).
The quick – or hasty – reaction to the September 11
attacks provides considerable evidence of learning, or of
political opportunism, as with the enactment of rather
stringent changes to criminal law enacted in the Patriot
Act but that have been more often used in run-of-the-
mill criminal cases than in prosecutions of terrorism.
This notion of political opportunism is not meant to
be cynical. Rather, it is a reflection of how ideas come
to the fore in both Cohen, March, and Olsen’s (1972)
‘garbage can’ model of organizational decision making,
as extended to the policy process in Kingdon’s ‘streams’
148 Thomas A. Birkland
Journal of Contingencies and Crisis Management
Volume 17 Number 3 September 2009 & 2009 Blackwell Publishing Ltd.
metaphor. After all, all groups have an ‘agenda’, which,
in American politics, at least, has come to sound like
something sinister (‘the liberal agenda’, ‘the right-wing
agenda’) but that really means the pre-existing goals
that groups seek to pursue. Clearly, if it is more
economical, in terms of political capital and the gen-
eration of public interest, to use an event as a way to
advance a group’s agenda, they will do so, such as when
environmental groups were able to use the Exxon Valdez
oil spill to advance claims that further development of
oil resources in Alaska would be environmentally
damaging (Birkland, 1997, Chapter 4).
Another type of reaction is one through which some
sort of learning (sometimes called ‘assessment’ or
‘evaluation’) process is begun, either within or outside
an agency, to assess what went well after an event, what
did not go well, and what should be improved in the
future. Such efforts, if done well, are designed to
understand the social, technological, and engineering
reasons for major failures that lead to disasters, such
as the multiple investigations of the levee failures
during hurricane Katrina conducted by expert investi-
gators. Others, who may not be as familiar with the
response as the experts, will develop ‘lessons learned’
documents that focus on particular aspects of their
concern that are based on secondary sources, and
that use the event as an exemplar. For example,
publications aimed at information technologists will
use an event to highlight lessons learned about the
physical security of computers, servers, and related
infrastructure, even though these ‘lessons’ were well
known before the event in question, and there is little
reason to believe that action as a result of these efforts
will be greater after the report than before. Indeed, we
might call all these lessons learned documents ‘lessons
observed’.
This is often well known to the participants in these
efforts, which is part of the investigatory process.
Leaders of investigative bodies pledge that their report
will not join a series of reports that ‘sit on a shelf and
collect dust’. Rather, their investigations will yield
tangible improvements in the way of policy and practical
change. Indeed, some members of the September 11
commission created the nongovernmental Public Dis-
course Project as a way to keep the recommendations
alive and in front of public officials, although this group
was disbanded at the end of 2005.
3.1. Potential patterns of ‘lessons learned’processes
There appear to be five broad patterns of ‘lessons
learned’ processes and documents:
� An event happens, and then change happens with
little or no effort devoted to learning from the
event. A major example is the USA Patriot Act,
which was enacted very soon after the September
11 attacks, without any real effort expended to see
whether the policy tools contained in that act
would really be the most effective in preventing
terrorist attacks.
� An event happens, and an investigation is under-
taken that is agency serving, is incomplete, or states
the obvious, without any evidence of a serious
attempt to learn. An example is the Executive
Office of the President’s Lessons Learned from
Katrina, the point of which is as much rhetoric as
it is real learning. Such reports simply hope to, in
Schattschneider’s (1975) terms, contain the scope
of conflict by creating the appearance of learning or
reform. Of course, there may well be some real
learning reflected in such reports, but their primary
function, ultimately, is public reassurance, not inter-
nal evaluation.
� An event happens, and an investigation is initiated,
which leads to policy change, but that policy change
cannot be linked to the investigation, or policy
changes without reference to the changes recom-
mended in the post-event investigation. For exam-
ple, there were many different attempts to
investigate September 11, but it is not clear
whether the creation of the Department of Home-
land Security was a direct outcome of these in-
vestigations, particularly given the thin evidence that
such an agency was really necessary (Tierney, 2005).
Indeed, DHS was created 2 days before the major
investigation – popularly known as the September
11 commission – was established. Its final report
was submitted in September 2004.
� An event happens, and a thorough and careful
investigation is initiated, but policy change does
not result. This may be because of cost, bureau-
cratic delay, political opposition, or any of the usual
reasons for political and policy stasis. For example,
the fruits of many NTSB investigations of airplane
crashes, including precursors to ValuJet 592, were
largely ignored for years by the Federal Aviation
Administration (Schiavo, 1997). The same is true for
aviation security problems before September 11,
where FAA moved very slowly in the face of what
was considered to be a growing threat (Birkland,
2004). However, we might still find the learning
process to be functional if the crisis was so anom-
alous that no intervention could improve policy
performance, such as the unforeseen ‘freak acci-
dent’, or if the remedy for the problem would
create more problems than the original problem
itself. For example, we know that some number of
people may be trapped in cars by seat belts in
accidents, and may perish in a fire if the car catches
fire. We also know that some very small fraction of
Disasters, Lessons Learned, and Fantasy Documents 149
& 2009 Blackwell Publishing Ltd.
Journal of Contingencies and Crisis Management
Volume 17 Number 3 September 2009
people who are vaccinated against diseases may
react badly to the vaccine, resulting in illness or
death. But we do not generally contemplate remov-
ing seat belts or halting vaccinations because the
broader social good these things do far outweighs
the small potential harms (while acknowledging, of
course, that the harms to those few injured indivi-
duals are not small).
� An event happens, and a thorough and careful
investigation is initiated, which leads to policy
change as a result of careful investigation, assess-
ment, and policy design. An example is the Colum-
bia Accident Investigation Board, which probed the
2003 space shuttle accident. There were changes at
NASA as a result of this report, including a much
closer inspection of heat shields and, in particular,
of potential damage to wings from falling foam
debris from the external fuel tank. However, one
must not make too much of ‘successful’ learning,
because these lessons can decay over time, as they
did between the loss of Challenger and Columbia. On
the other hand, the second shuttle accident has
led to fundamental rethinking about spaceship de-
sign, with new craft being simplified and designed to
put the crew far forward of the dangerous fuel
tanks; this focus on safety and survivability is a
function of double-loop learning. However, many
careful investigations yield single-loop learning that
does yield operational and regulatory change with-
out being elevated to the legislative level. An
example is the NTSB’s and the FAA’s investigation
of a series of rudder deflection incidents that
included the crash of US Airways flight 427 near
Pittsburgh in 1994. This investigation ultimately led
to the discovery and remedy of a design flaw with
the mechanism that controlled the Boeing 737-300
rudder (see http://www.ntsb.gov/events/usair427/
items.htm). Indeed, the NTSB’s work on aviation
accidents is considered a model of learning from
thousands of minor to major incidents that accu-
mulate into a vast body of operational knowledge
(Perrow, 1999).
The first four of these examples falls into a class I call
‘fantasy learning’ that generates ‘fantasy lessons learned
documents’, although the fourth example might be
more a function of bureaucratic delay rather than
of rhetoric. Only one of these scenarios – the fifth –
is an example of sound instrumental learning. While
this sort of rational, experience-, and evidence-based
learning is considered by the public and many actors
to be a desirable outcome of such events, and describes
what we might consider the classical model of learning,
this sort of learning is rare. There are many reasons,
then, for the production of fantasy lessons learned
documents:
4. A model of event-related policychange
The logic model in Figure 1 depicts the ideal process of
event-related learning, which can be used to test the
patterns of lessons learned processes. In this model, if
certain actions occur at points after a focusing event
occurs, learning becomes more likely, and policy change
as a result of this learning becomes more likely. This
model also suggests that after an event, it is possible for
learning without policy change to occur after one event,
or for policy change to result from mimicking or
‘superstitious’ learning. This learning is the result of
pressure to ‘do something’ after an event, and where
issuing a ‘lessons learned’ document is taken to be
evidence of at least the beginning of an effort to tackle
the failures revealed by the event. Finally, the model
acknowledges that not every event will lead to policy
change, but that events may contribute to a base of
experience that may promote learning from subsequent
events as knowledge accumulates, as noted in the
feedback arrow. In other words, not all events do
involve explicit acknowledgement of lesson learning.
In this model, I operationalize learning in this way:
first, I adopt Busenberg’s process-based definition but
stipulate that focusing events, consistent with Kingdon’s
streams metaphor, and Cohen March and Olsen’s ‘gar-
bage can’ (Cohen et al., 1972) model, on which Kingdon
relies, that definition of learning as ‘a process in which
individuals apply new information and ideas to policy
decisions’. However, I modify this definition slightly to
define learning as a process in which individuals apply
combined new information that may be revealed by a
disaster with and ideas, or new and preexisting information
and ideas elevated on the agenda by a recent event, to actual
policy change, policy decisions. This redefinition takes
into account two factors: the ebb and flow of ideas on
the agenda and the accumulation of ideas over time, even
as those ideas are not uniformly translated into policy.
I do not claim to be able to measure ‘learning’ at the
individual level based on behavioural or cognitive science.
Rather, I focus on the apparent lessons of these events,
and ask whether it appears that the clear lessons of these
events have been learned, as reflected in the policy-
making process. In particular, we can say that there is
prima facie evidence of learning if policy changes in a way
that is reasonably likely to mitigate the problem revealed by
the focusing event. This operationalization of learning
cedes a great deal of judgement to the researcher making
the claim of learning. This is why clear criteria and coding
frames are necessary to any detailed study of learning.
4.1. Drivers of the learning process
What is the motive force that advances the learning
process? I identify three drivers of this process, all of
150 Thomas A. Birkland
Journal of Contingencies and Crisis Management
Volume 17 Number 3 September 2009 & 2009 Blackwell Publishing Ltd.
which can either promote learning or lead to dysfunc-
tional learning. The first driver is the desire to learn,
quickly, why a bad thing happened so as to prevent
its recurrence. These pressures create hasty attempts
to learn from events, which can induce pre-mature
attribution of causes, such as the early claims by Jack
Kallstrom, the FBI’s New York bureau chief, that TWA
flight 800 was brought down by a bomb in 1996; it
turned out that a careful analysis found that the plane
exploded due to an abundance of explosive vapours in a
fuel tank. The news media are notoriously prone to
both warning against speculation and then speculating
about the causes of airplane crashes, sometimes in the
same story.
Self-interest is not simply about attempting to inocu-
late an agency or a group against criticism. The mirror
image of the self-promoting ‘lessons learned’ process is
a wildly critical effort that seeks to find fault with
everything that everyone did in an event. Few reports
are this critical, but the legislative branch is often
tempted, for partisan or institutional reasons, to focus
on failures and ignore successes. Sometimes, these
failures are overstated or personalized, as in Congress’s
grilling of former FEMA director Michael Brown after
An EventOccurs
IncreasedAgenda
Attention
GroupMobilization’
Discussionof Ideas
New PoliciesAdopted
Yes
No
PossibleInstrumental orSocial Learning
Yes
Yes
Yes
No
No
No
Little or nolearning
Little or nolearning
New PoliciesAdopted?
Yes
Possible superstitiouslearning or mimicking
Possible politicalor social learning,could be appliedto future events
Possibleaccumulation of
learning for futurepolicy making
Accumulatedexperience from
prior events
No
Figure 1. A Model of Event-Related Policy Learning.
Disasters, Lessons Learned, and Fantasy Documents 151
& 2009 Blackwell Publishing Ltd.
Journal of Contingencies and Crisis Management
Volume 17 Number 3 September 2009
hurricane Katrina, which attempted to attribute
many of the problems encountered in hurricane Katrina
to one person’s purported incompetence, not to
systemic failures.
On the other hand, the political and time pressure
created by a crisis may create a sense of purpose and
urgency that would not otherwise exist without the
crisis having happened. The investigations of the losses
of space shuttles Challenger and Columbia were driven
by the very fact that they led to loss of life (and, less
publicly, by the significant costs of losing these space-
craft). Urgency can therefore be a productive or a
distorting force.
A second driver of the learning process is individuals’
or groups’ self-interest. The choice to call a document a
‘lessons learned’ document can be strategic and rheto-
rical, and is revealed by the policy prescriptions to which
the report leads. For example, the American Highway
Users Alliance commissioned a study (American High-
way Users Alliance, 2006) to demonstrate the ‘need’ for
better evacuation planning using private automobiles and
over the road buses to allow entire cities to evacuate
because of what was ‘learned’ about the ‘failed evacua-
tion’ of New Orleans. While this study was triggered by
hurricane Katrina, this study was based almost entirely
on industry self-interest, was methodologically deficient,
and failed to take into account the largely successful
evacuation of New Orleans and its environs (Roig-
Franzia & Hsu, 2005; Wolshon et al., 2006; Derthick,
2007). The report’s credibility was further undermined
by its authorship by a consultant with a strong pro-
automobile, anti-transit, and anti-planning bias.
A third driver is the human tendency, under bounded
rationality, to attempt to find simple or monocausal
explanations for very complex social and political pheno-
mena. Focusing on one or a few aspects of a disaster will
not often get to the heart of the problem. For example,
the concentration of attention on New Orleanians’
choices to live in the parts of the city resting below
sea level seemed to create a causal story that focused on
the ‘poor decisions’ of the people who live there, which
is another version of ‘operator error’ rather than of
systemic error. The implicit lesson is that people should
be discouraged from living in vulnerable areas, but this
construction of vulnerability fails to account for a wide
range of things that create vulnerability. These include
complex socioeconomic and demographic factors, the
political economy of the region, the physical landform,
the roles of other actors (the Orleans Parish Levee
Board, the Corps of Engineers, the city and state
governments), and so on (Cooper & Block, 2006). The
blaming of individuals for the failure, such as the afore-
mentioned criticisms of Michael Brown, is yet another
example of monocausal attribution, as are ‘operator
error’ causes in complex systems accidents, such as
aircraft or nuclear power (Perrow, 1999).
4.2. Propositions about event-driven learning
The goal of the logic model is to generate propositions
about after disaster learning. These propositions also
suggest the data needed to understand the phenom-
enon of interest. I do not claim that these are hypoth-
eses, because further model development and theory
building is required. But I advance these as guidance for
future research.
The first proposition is that a few events will gain the
most attention. The distribution of damage and deaths in
disasters and accidents is not statistically normal;
rather, the distribution has a long ‘tail’, where a large
number of relatively small events garner little attention,
and a few events gain a great deal of attention. For
example, many tropical storms or hurricanes that can
strike the nation during the hurricane season, but only
the very few largest storms, on the scale of hurricanes
Katrina or Andrew, receive the most attention and can
have the greatest influence on learning. Smaller inci-
dents do not gain attention because they place less
strain on existing organizations and policies; in other
words, they are ‘routine’ disasters to organizations
designed to respond to such events. Hurricane Katrina
receive more attention than did all four of the hurri-
canes that struck Florida in 2004 because the response
to the Florida hurricanes was generally perceived as
adequate, and because no individual storm was cata-
strophic, while Katrina was a catastrophe that over-
whelmed the national emergency management system.
The disaster–catastrophe distinction is important,
because we can think of a disaster as affecting a
relatively small area whose emergency response may
be strained, but not overwhelmed, while a catastrophe
entirely overwhelms the ability of a community or its
region to respond (Quarantelli, 2005), as was evident in
hurricane Katrina. This distinction is important because
it reflects the greater scale of the catastrophe. In
English, this distinction is much more pronounced
than in, for example, French, where catastrophe naturelle
usually translates to ‘natural disaster’ in English.
The second proposition is that most, if not all,
participants in a policy domain want to address or
solve the problems revealed by a focusing event, but
that the proposed solutions will likely vary with the
interests and motivations of the various participants.
This reflects the idea that nearly all participants in a
domain are goal oriented (Jones, 2001). No legitimate
actor in any policy domain wants to see planes hijacked
or people displaced due to natural disasters. But the
policy instruments with, which problems will be pre-
vented or mitigated will differ from participant to
participant in the policy process, because the depiction
of how problems come to be, and therefore solved, will
be different based on each participant’s ideological and
organizational commitments.
152 Thomas A. Birkland
Journal of Contingencies and Crisis Management
Volume 17 Number 3 September 2009 & 2009 Blackwell Publishing Ltd.
The third proposition, related to the second, is that
group mobilization is linked in time to a particular focusing
event. In particular, the activities of groups – or the
representatives of such groups – will become more
evident in news accounts of the crisis or disaster as it
unfolds. In congressional hearings (or parliamentary
inquiries), particular groups’ representatives will be
heard from more often.
The fourth proposition is that group mobilization will be
accompanied by an increased discussion of policy ideas.
These will include theories about the causes and poten-
tial solutions of the problem, and, as such, are primarily
social and instrumental policy learning matters. I assume,
therefore, that events drive group mobilization, which
drives the discussion of policy ideas, again consistent
with the ‘garbage can’ model of decision making (Cohen
et al., 1972). Evidence of political learning may also exist,
but such evidence may be less apparent, given that this
learning happens internally within organizations in the
policy domain or advocacy coalitions. In any case, policy
learning is much less likely without the mobilization of
tangible ideas, and ideas are unlikely to come to the fore
without some sort of group mobilization.
Thus, the fifth proposition is that there is a relationship
between ideas and policy change. In particular, change is
more likely when there are ideas triggered because of
events, compared with when there are no ideas gener-
ated by an event or elevated to a higher position on the
agenda. Policy change can occur without ideas, but we
can assume that such policy change does not happen
because of careful debate of ideas and therefore does
not result from learning; instead, it is mimicking or
copying without learning (May, 1992). Table 1 shows the
types of evidence one would use to illustrate learning as
conceptualized in these propositions.
The sixth proposition is that it is possible for the
lessons learned to decay over time. While policy change
may result from an event, the time that intervenes
between one focusing event and another, and the
demands placed on policy makers in that intervening
period, may cause participants in the policy process to
‘forget’ the lessons that they learned. The effect of
hurricane Katrina, and the fumbled federal, state, and
local response to the event, suggested that the putative
lessons of hurricane Andrew were not fully learned,
forgotten over time, or were influenced by the inter-
action between the natural hazards and the ‘homeland
security’ domains. Kingdon calls these interactions
between policy domains ‘spillovers’, and such spillovers
can theoretically reinforce learning, or can retard it.
The focus on homeland security had a corrosive
influence on the nation’s pre-paredness for natural
disasters (Tierney, 2005). None of this is to suggest a
normative claim that lessons should not decay over
time; rather, it is to acknowledge that any lesson will
necessarily decay unless it is fully institutionalized into
the law, from legislation through regulation to the
standard operating procedures of regulations.
5. Interim observations on the modeland propositions
This article started with the idea of the lessons learned
document as a ‘fantasy document’. The paper then
proceeded to explain a model of crisis-spurred policy
learning, including its main drivers and key propositions
that derive from the model. Clearly, the entire concept
of ‘fantasy learning’ is broader than the actual document
itself. Rather, I describe a process where the production
of a document is a final or even an interim step along a
much longer timeline, where the document might signal
the end of a period of significant reflection, or may
mark the beginning of further controversy over what
was claimed to have been learned. I focus on the
document as a key feature of the analysis because the
thinking that often goes into such documents reflects
both the functional and the dysfunctional features of the
learning process I outline here. The functional features
include improved policy that yields improved perfor-
mance; the dysfunctional features involve features that
impede learning, or that would, for whatever reason,
prevent what was learned from being put into practice.
Of course, by contrasting ‘functional’ and ‘dysfunc-
tional’ aspects of learning, I appear to adopt a function-
alist perspective on the entire policy process. But
scholars of public policy have long known that most
policy problems are socially constructed and are em-
bedded in long-standing ideas, norms, and practices.
Framing of problems and their solutions is a key part of
this process. It is important to acknowledge that the
learning described in this article is about lessons that
may already be well known, or that were ‘learned
Table 1. Typical Evidence of Learning in the Policy Process
Organization orinstitution Evidence of learning
News media Stories about the problemChanges in the nature of news coverage(people quoted, substance of news cover-age).
Interest groups Change in appearances at congressionalhearings.Increased attention from news media(generated by the group).
Congress Legislative change.Change in the substance of debate.Change in the topic areas of hearings.
Regulatory andimplementingagencies
Issuance of new and proposed regulations.Change in the nature and substance of theregulations being issued. Change in proce-dures and in the interpretation and imple-mentation of statutes and regulations.
Disasters, Lessons Learned, and Fantasy Documents 153
& 2009 Blackwell Publishing Ltd.
Journal of Contingencies and Crisis Management
Volume 17 Number 3 September 2009
before’ but that become dormant between events, and
the very nature of the lesson-learning process will
depend on how the original policy failure – the problem
itself – is framed. Considerable contention can result
when there are different interpretations of the pro-
blem, because these different interpretations and claims
will greatly influence the claims about what the ‘lessons’
should be. In such an environment, even the claim of
‘fantasy’ learning is contested, because, after all, who is
to say that the learning process is ‘real’ vs. ‘fantastic’?
This paper suggests, however, that there are important
distinctions between learning that is functional in the
sense that it yields policy change and improvement, and
dysfunctional ‘fantasy’ learning that may be driven by
poor causal theory or by narrow self-interest.
In working through this model of policy learning, and
accounting for the special conditions of learning from
crises and disasters, there are important avenues for
future research and for refining this model. After all, it
is a tall order to expect that a policy network will
experience a disaster, will take the necessary steps to
learn from it, and then will put those lessons into effect.
The first issue deserving of attention is the combined
question of time pressures and the overwhelming
publicity that surrounds crises and disasters. Indeed,
the most relevant feature of large disasters is that they
are so huge that their harmful nature is immediately
clear to all in the disaster area, and to those who learn
of the disaster through the news. Containing the scope
and scale of the disaster is the main goal of decision
makers during a crisis, but they must work very quickly
to achieve this end. They do not have a great deal of
time to be reflective and, instead, must often improvise
to find good interim solutions to problems that were
unanticipated, or to problems that cannot be amelio-
rated through standard operating procedures in routine
times, or even routine emergencies such as a small
chemical spill or a relatively minor hurricane.
The second issue is the question of single- vs. double-
loop learning. Single-loop learning is generally learning
about tactics or operations, and is therefore not a key
feature of my model of the policy learning process. I
am more concerned with broader strategic learning
about the usefulness and appropriateness of policy
tools. These policy tools are presumed to have failed
in a crisis, and the crisis is, therefore, an opportunity to
learn and to improve our knowledge of problem solving
at the instrumental level (the policy tool) or at the
social level, involving better understandings of cause
and effect relationships, rhetoric, or the tractability of
public problems. But the line between the types of
learning is blurry, at best. Learning about policy tools,
even at the legislative level, certainly invokes opera-
tional issues. The learning I am most interested in this
paper therefore suggests some sort of fundamental
rethinking about policy besides its operational aspects.
This is why I put the ‘fantasy document’ at the start of
my investigation of learning in this article; such docu-
ments are the end point of an ongoing process. But the
real point is less the document than it is the process
that yielded the ultimate document. We might there-
fore wish to test the process from its outset, by asking
whether the process was an ‘honest’ attempt to learn,
or whether the process was a public relations activity
or a ‘whitewash’ intended to burnish the image of an
organization, or to absolve it of responsibilities for
failures. One might approach this question by finding
out whose office was ultimately responsible for compil-
ing and disseminating any ‘lessons learned’. If we learn
that the public relations staff developed such reports,
one might approach the entire process much more
sceptically than if one knew that the report was created
by a serious internal effort, and external review body,
or some combination of the two.
Indeed, this points out a flaw in the idea that there is
‘one’ lessons learned document. Future research should
look into the range of ‘lessons’ documents that are
produced after a crisis or a disaster. These include
anything from changes to standard operating proce-
dures to major statutory changes, as well as internal
reports and analyses. There may be some divergence
between the public face of an organization and its
private deliberations, particularly under conditions of
extreme attention and time pressure.
6. Conclusion
To call a ‘lessons learned’ document a fantasy document
is to call the entire process by which the document was
created a fantasy exercise. This is not true, of course,
in all cases – there have been many earnest efforts
to improve performance after a crisis or a disaster,
and some – but by no means all – of these efforts
have improved performance. But, in many cases, when
viewed from a political perspective, learning processes
are often not ‘serious’ in the sense that they are
intended to extract lessons from experience and apply
them to current and future problems. Instead, many of
these documents and the processes that create them
are mere reflections of a group’s or interest’s preferred
social construction of a problem and its ‘target popula-
tions’. Often, these groups will resist serious lesson-
learning processes by either resisting the creation of
such investigations, or will, once the investigation is
complete, deny the lessons on cost, feasibility, or other
grounds, or will simply ignore them. For these reasons,
learning is not as common as one might think, even if
the participants in these processes sincerely believe
that the process in which they are engaged is intended
to learn something. Many of these participants learn
that they have to communicate ex cathedra if their ideas
154 Thomas A. Birkland
Journal of Contingencies and Crisis Management
Volume 17 Number 3 September 2009 & 2009 Blackwell Publishing Ltd.
are to gain attention in future policy debates. More
often, these processes simply result in reports that fail
to address the real problems revealed by an event or a
series of events. The challenge for democracies is to
create the sort of public pressure necessary to make
learning processes more realistic and responsive to the
problems and to the needs of the organizations, com-
munities, regions, and nations in which these events
occur. Because many political systems contain features
that prevent rather than promote policy change, such
learning efforts are doubly challenged, and a great deal of
energy is necessary to overcome systemic inertia. But, in
some cases, learning can exist, and we can ‘learn’ from
these processes how to structure organizations and
policy systems that bring serious learning to the fore.
Note
1. In sharpening the definition, I acknowledge that I also
narrowed the definition substantially, thereby ignoring the
influence of personal experience among decision makers,
among other factors, as type of focusing events. There is
likely some sort of typology of focusing events, which is
beyond the scope of this paper.
References
Airline Industry Information (2001), ‘Cockpit Doors will
Remain Locked Throughout All Canadian Flights, Airline
Industry Information,’ 18 September, via LexisNexis.
Air Safety Week (2000), ‘Bar the Door, Air Safety Week,’ 14
(12), via LexisNexis.
American Highway Users Alliance (2006), Emergency Evacuation
Report Card 2006, American Highway Users Alliance,
Washington, DC, http://www.highways.org/pdfs/evacuation_
report_card2006.pdf (accessed 1 May 2008).
Argyris, C. and Schon, D. (1996), Organizational Learning,
Addison-Wesley, Reading.
Baumgartner, F.R. and Jones, B.D. (1993), Agendas and Instabil-
ity in American Politics, University of Chicago Press, Chicago.
Birkland, T.A. (1997), After Disaster: Agenda Setting, Public
Policy and Focusing Events, Georgetown University Press,
Washington, DC.
Birkland, T.A. (1998), ‘Focusing Events, Mobilization, and
Agenda Setting’, Journal of Public Policy, Volume 18, Number
3, pp. 53–74.
Birkland, T.A. (2004), ‘Learning and Policy Improvement After
Disaster: The Case of Aviation Security’, American Beha-
vioral Scientist, Volume 48, Number 3, pp. 341–364.
Birkland, T.A. (2006), Lessons of Disaster, Georgetown
University Press, Washington, DC.
Busenberg, G.J. (2001), ‘Learning in Organizations and Public
Policy’, Journal of Public Policy, Volume 21, Number 2, pp.
173–189.
Clarke, L.B. (1999), Mission Improbable: Using Fantasy Documents
to Tame Disaster, University of Chicago Press, Chicago.
Cobb, R.W. and Elder, C.D. (1983), Participation in American
Politics: The Dynamics of Agenda-Building, Johns Hopkins
University Press, Baltimore.
Cobb, R.W. and Primo, D.M. (2003), The Plane Truth: Airline
Crashes, the Media, and Transportation Policy, Brookings
Institution, Washington.
Cohen, M.D., March, J.G. and Olsen, J.P. (1972), ‘A Garbage
Can Model of Organizational Choice’, Administrative Science
Quarterly, Volume 17, pp. 1–25.
Cooper, C. and Block, R. (2006), Disaster: Hurricane Katrina and
the failure of Homeland Security, Times Books, New York.
Derthick, M. (2007), ‘Where Federalism Didn’t Fail’, Public
Administration Review, Volume 67, Number s1, pp. 36–47.
Haider-Markel, D.P. and Joslyn, M.R. (2001), ‘Gun Policy,
Opinion, Tragedy and Blame Attribution: The Conditional
Influence of Issue Frames’, Journal of Politics, Volume 63,
Number 2, pp. 520–543.
Hilgartner, J. and Bosk, C. (1988), ‘The Rise and Fall of Social
Problems: A Public Arenas Model’, American Journal of
Sociology, Volume 94, Number 1, pp. 53–78.
Jones, B.D. (2001), Politics and the Architecture of Choice:
Bounded Rationality and Governance, University of Chicago
Press, Chicago.
Kingdon, J.W. (1995), Agendas, Alternatives and Public Policies,
Harper Collins: New York.
Larkin, R.W. (2007), Comprehending Columbine, Temple Uni-
versity Press, Philadelphia.
Lawrence, R.G. and Birkland, T.A. (2004), ‘Guns, Hollywood,
and Criminal Justice: Defining the School Shootings Pro-
blem across Public Arenas’, Social Science Quarterly, Volume
85, Number 5, pp. 1193–1207.
May, P.J. (1992), ‘Policy Learning and Failure’, Journal of Public
Policy, Volume 12, Number 4, pp. 331–354.
Naik, G., Sprothen, V. and Crawford, D. (2005), Scandinavians’
Sense of Order is Shattered by Asian Disaster, Wall Street
Journal, New York, 7 January, via LexisNexis.
Neustadt, R.E. and May, E.R. (1986), Thinking in Time: The Uses
of History for Decision Makers, Free Press, New York.
Perrow, C. (1999), Normal Accidents: Living with High-Risk
Technologies, Princeton University Press, Princeton, NJ.
Quarantelli, E.L. (2005), Catastrophes are different from disasters:
some implications for crisis planning and managing drawn from
Katrina, http://understandingkatrina.ssrc.org/Quarantelli
(accessed 21 October 2005).
Richfield, P. (2000), ‘Secret Knock Not So Secret, Business
and Commercial Aviation,’ 36 August, via LexisNexis.
Roig-Franzia, M. and Hsu, S. (2005), Many Evacuated, but
Thousands Still Waiting, White House Shifts Blame to State
and Local Officials, Washington Post, Washington, DC,
4 September, A1, via LexisNexis.
Rubin, C.B., Cumming, W.B., Tanali, I.R. and Birkland, T.A.
(2003), Major Terrorism Events and their U.S. Outcomes (1988–
2001), Natural Hazards Research and Applications Informa-
tion Center, Institute of Behavioral Science, University of
Colorado, Boulder CO, http://www.colorado.edu/hazards/
wp/wp107/wp107.html (accessed 1 December 2008).
Schattschneider, E.E. (1975), The Semisovereign People, The
Dryden Press, Hinsdale, IL.
Schiavo, M. (1997), Flying Blind, Flying Safe, Avon Books,
New York.
Disasters, Lessons Learned, and Fantasy Documents 155
& 2009 Blackwell Publishing Ltd.
Journal of Contingencies and Crisis Management
Volume 17 Number 3 September 2009
Stromback, J. and Nord, L.W. (2006), ‘Mismanagement,
Mistrust and Missed Opportunities: A Study of the 2004
Tsunami and Swedish Political Communication’, Media,
Culture and Society, Volume 28, Number 5, pp. 789–800.
Tierney, K.J. (2005), The red pill. http://understandingkatrina.
ssrc.org/Tierney/ (accessed 12 November, 2008).
Widfeldt, A. (2007), ‘Sweden’, European Journal of Political
Research, Volume 46, Number 7/8, pp. 1118–1126.
Wolshon, B., Catarella-Michel, A. and Lambert, L. (2006),
‘Louisiana Highway Evacuation Plan for Hurricane Katrina:
Proactive Management of a Regional Evacuation’, Journal of
Transportation Engineering, Volume 132, Number 1, pp.
1–10.
World Airline News (2001), ‘International Bodies Take Action
To Enhance Security Programs’, World Airline News, 14
September, via LexisNexis.
156 Thomas A. Birkland
Journal of Contingencies and Crisis Management
Volume 17 Number 3 September 2009 & 2009 Blackwell Publishing Ltd.